An Investigation On The Problems Of EMD And AR Based Short Time Prediction Of Ship Motions

The shortest time prediction of ship motions means adopting some methods to predict ship motion over an interval several seconds according to the given information of ship motion and ocean environment.This paper adopts the analysis method of time series to predict the ship motions. Further research is done based on Wang An-cun's paper. The basic thought of this method is decomposing non stationary and non linear data of ship motions into several intrinsic mode functions (IMF) using the Empirical Mode Decomposition(EMD) method and formulating Auto-regression model (AR) of every IMF respectively to predict, and finally adding all the prediction results. The contents include four parts:First, the Empirical Mode Decomposition (EMD) is an effective method of analyzing the non-linear and non-stationary process data. However, the extension of boundary and extreme point are still needed to be investigated. The paper made calculations based on this new signal analysis method by FORTRAN program. Further research combing the requirement of the shortest time prediction is done. A method to extend the data at the boundary and an effective method to select the extreme points in the data is presented.Second, the comparative study on parameter estimate methods and definite orders in Auto-regression (AR) is done using FORTRAN program. The paper also compares of the influence of different length of signal data on AR prediction.Third, the paper adopts EMD combing AR to predict and study the influence on prediction results due to different sample data.Finally, in order to consider the signal non-linearity, we get the parameter of the SETAR model according to local area search method and predict ship motions.